betaFeedback

WorksPhysics and Astronomy

Fokker-Planck Model Based Central Moment Lattice Boltzmann Method for Effective Simulations of Thermal Convective Flows

Abstract

The Fokker-Planck (FP) equation represents the drift-diffusive processes in kinetic models. It can also be regarded as a model for the collision integral of the Boltzmann-type equation to represent thermo-hydrodynamic processes in fluids. The lattice Boltzmann method (LBM) is a drastically simplified discretization of the Boltzmann equation. We construct two new FP-based LBMs, one for recovering the Navier-Stokes equations and the other for simulating the energy equation, where, in each case, the effect of collisions is represented as relaxations of different central moments to their respective attractors. Such attractors are obtained by matching the changes in various discrete central moments under collision with the continuous central moments prescribed by the FP model. As such, the resulting central moment attractors depend on the lower order moments and the diffusion tensor parameters and significantly differ from those based on the Maxwell distribution. The diffusion tensor parameters for evolving higher moments in simulating fluid motions at relatively low viscosities are chosen based on a renormalization principle. The use of such central moment formulations in modeling the collision step offers significant improvements in numerical stability, especially for simulations of thermal convective flows with a wide range of variations in the transport coefficients. We develop new FP central moment LBMs for thermo-hydrodynamics in both two- and three-dimensions and demonstrate the ability of our approach to accurately simulate various cases involving thermal convective buoyancy-driven flows, especially at high Rayleigh numbers. Moreover, we show significant improvements in numerical stability of our FP central moment LBMs when compared to other existing central moment LBMs using the Maxwell distribution in achieving higher Peclet numbers for mixed convection flows.

Introduction

Fluid motions coupled with thermal energy transport arise in a wide range of situations and scales, from atmospheric convection and heat transfer in process equipment and power generation devices to bioengineering systems and microfluidic applications. Macroscopically, the dynamics of fluids is represented by the Navier-Stokes equations (NSE) while the thermal transport is generally modeled using a convective-diffusion equation (CDE) for the temperature field, sometimes classified as having a combination of hyperbolic and parabolic behavior [1]. As such, the solution of the CDEs is needed not only in understanding or making predictions in heat transfer applications but also in the transport of passive scalar fields involving chemical species, in the interface capturing in multiphase flows using phase field models, and in modeling of surfactants (for recent examples, see [2, 3, 4, 5] that involve a combination of such cases and other related applications [6, 7]). Of specific interest to this work is the computation of buoyancy-driven thermal convective flows [8], which involve a two-way coupling between the solution of NSE and CDE for the temperature field.

Obtaining solutions to such a set of coupled partial differential equations for realistic situations using analytical methods is infeasible. For that reason, the development and use of numerical techniques plays a fundamental role in simulating and studying the complex behavior of thermal transport phenomena in fluids. Traditionally, methods such as the finite difference, finite element, and especially finite volume methods (see e.g., [9, 10]) have been employed in discretizing the macroscopic governing equations; they are often used in conjunction with other techniques such as multigrid methods (see e.g., [11, 12, 13]) for convergence acceleration and have been investigated previously in detail and are quite commonplace. On the other hand, the focus of this paper is on the development of a novel formulation of lattice Boltzmann (LB) method originating from kinetic theory and in particular as a dramatic simplification of the continuous Boltzmann equation. Much of the recent interest in such methods arise from their computational advantages including their effective implementation on modern parallel computing architectures due the local nature of the algorithms, ease of implementation of boundary conditions using Cartesian grids, and their natural ability to use more fundamental mesocopic models based on kinetic theory to represent complex flow physics [14, 15, 16]. Moreover, the LB schemes are flexible in that it can be combined with concepts used in conventional approaches such as unstructured grids via finite volume methods to handle flows in complex geometries (see e.g., [17, 18, 19, 20]).

As such, the LB method is an efficient numerical approach for solving the mildly compressible NSE, as well as for convection-diffusion type equations, and many other models for fluid dynamics with multiphysics effects. The underlying idea in the LB method is that a simplification of the continuous Boltzmann equation [21] is used to track the density distribution function of fluid particles, whose motion has been restricted to discrete velocity directions, called lattice links, as they move around and collide. More specifically, the collision process is modeled as relaxations of those distribution functions towards their ‘equilibrium’ states, typically represented as a discretized form of the Maxwell distribution, while accounting for any relevant conservation laws. Following this, the distributions are then streamed or advected along the discrete lattice links, and the fluid dynamical variables are then computed via taking the moments of the distribution functions with respect to the powers of the particle velocity components. The LB methods have been applied to simulate a wide range of problems involving multiphase and multicomponent flows, particulate suspension flows, thermal flows, turbulence, combustion, magnetohydrodynamics and aeroacoustics (see e.g., [22, 23, 24, 25] for some recent reviews).

Modeling the collision process in the LB method is a decisive feature used to extend the physics of flow characteristics in the simulation and, more generally, the numerical stability; to that end, many collision models have been developed. The first collision model constructed is known as the single relaxation time model (SRT) [26], setting the stage for modeling collision as a relaxation process using a single characteristic time scale despite the fact the underlying kinetic phenomena can be multiscale in nature; however, it was found to have serious limitations in terms of numerical stability. To alleviate this issue, the multiple relaxation time (MRT) collision model was developed which relaxes different bare (or raw) moments of the distribution functions at different rates [27, 28], which was found to have significant improvements in numerical stability compared to the SRT model. Moreover, as a further improvement, the central moment collision model, which is also referred to as the cascaded collision model, was then constructed [29], which involves performing collisions using central moments. Here, the central moments are formed by taking moments of the distribution functions with respect to the powers of the components of the particle velocity being shifted by the bulk fluid velocity. Such an approach entails the use of additional transformations between raw moments and central moments when performing the collision step; as such, the latter involves the relaxations of central moments to the equilibria defined by the central moments of the Maxwell distribution and this approach resulted in significantly enhancing the numerical stability when compared with the MRT model while naturally maintaining Galilean invariance (see Ref. [30] for a comparative numerical study of these collision models). However, it was found to have numerical hyperviscosity artifacts when such relaxations are done at different rates for different central moments in simulating high Reynolds number flows. In order to address this issue, a factorized central moment collision model was introduced [31], which exploits the factorization property of the Maxwell distribution to construct the attractors in an ad hoc manner. Another solution to this issue was provided by developing a collision model that relaxes the cumulants of the distribution function toward their attractors [32], which involves an additional transformations for mappings between central moments to cumulants resulting in added complexity.

More recently, we proposed a new approach to the collision modeling in LB methods that involves using the central moments of the Fokker–Planck (FP) equation to simulate fluid dynamics at high Reynolds numbers, which was shown to naturally resolve the numerical hyperviscosity issues while also showing superior numerical stability properties resembling those observed in the cumulant collision model, but using fewer transformations and simpler formulations [33]. As such, the FP equation is an important formulation in statistical mechanics for evolving the density distribution function of particles under drift-diffusive processes in a variety of applications (see e.g., [34, 35, 36, 37]); it is equivalent to Langevin equation, a stochastic differential equation originally derived to model Brownian dynamics (see e.g., [38] for a pedagogical discussion on these topics) and is sometimes also referred to as the Kolmogorov forward equation or as the Smoluchowski equation. It has been used to model the collision term of the Boltzmann equation by Lebowitz et al. [39] to represent fluid dynamics and it has been shown to be a rigorous approximation to the Boltzmann’s collision integral under grazing collisions and a diffusion limit [40]. As shown in our recent work, a central moment formulation of the FP collision model results in a local and algebraic operator that is naturally suitable for algorithmic implementation in LB methods for effective simulations of fluid dynamics [33]. One key feature of this FP-based central moment LBM is that the resulting attractors contain products of lower order central moments and a diffusion tensor parameter and latter is specified as fixed points of the renormalization group equation representing the variations of the second order moments under collision when there is scale separation in the relaxation rates of kinetic processes. These considerations were shown to be naturally circumventing the hyperviscosity artifacts while also improving the numerical stability significantly when simulating flows at relatively very low physical viscosities or high Reynolds numbers [33] unlike the Maxwellian based central moment LB formulations which are susceptible to numerical issues. One of the major objectives of this paper is to extend such FP-based central moment LB schemes to simulate thermo-hydrodynamics, i.e., specifically to simulate the energy transport equation in addition to the NSE as done in the previous work. Based on the above discussion, the various key features and numerical attributes associated with different collision models in the LBM are summarized and presented in Table 1.

Collision model in LBMFeatureStability characteristicsHyperviscosity effects
SRT-LBMRelaxes distribution functions to their equilibria at same rateLowNo
MRT-LBMRelaxes raw moments to their equilibria at different ratesModerateYes
MCM-LBMRelaxes central moments to Maxwellian equilibria at different ratesHighYes
FPC-LBMRelaxes central moments to FP-based Markovian attractors at different rates and using a renormalization principleVery highNo
Cumulant LBMRelaxes cumulants to their equilibria at different ratesVery highNo

Table 1. Comparative summary of the features, numerical stability characteristics for simulations of flows at high Reynolds numbers, and the emergence of numerical hyperviscosity effects for flows at very low physical viscosities with different collision models in LBM: single relaxation time (SRT)-LBM, multiple relaxation times (MRT)-LBM, Maxwellian-central moments (MCM)-LBM, Fokker-Planck central moment (FPC)-LBM, and cumulant LBM.

Before discussing it in more detail, let us briefly review the prior work on developing LB schemes for flows with heat transfer. As such, they can be classified into the following three major types – 1) multispeed LB schemes using a single distribution function to compute both the hydrodynamic and temperature fields, 2) the hybrid approaches using a LB scheme for fluid motions and a finite difference method the energy transport, and 3) the double distribution functions (DDF)-based formulations that use one LB scheme for the NSE and another one for the energy equation (see e.g., [25, 41] for further details). Out of these, the DDF approach has become the most popular due to stability and efficiency considerations. Earlier investigations adopted the SRT collision model in the DDF-LB schemes (see [42, 43]), which was followed by the use of MRT models to simulate thermal convective flows in this framework [44, 45]. More recently, the central moments-based collision models were utilized to construct DDF-LB schemes in two-dimensions (2D) [46, 47, 48], in three-dimensions (3D) [49], and in axisymmetric geometries [50] with improved stability characteristics when compared to the methods based on SRT and MRT collision models. It should be noted that all such central moments-based formulations used the Maxwell distribution function, which depend only on the conserved hydrodynamic fields, in constructing the collision step.

As further improvements over these developments, in this paper, we will develop new DDF-based LB schemes using the FP central moment (FPC) collision operators in both 2D and 3D. In 2D, we will use the two-dimensional, nine velocity (D2Q9) lattice to construct the FPC-LBMs for both fluid motions and energy transport, and in 3D we will consider the three-dimensional, twenty-seven velocity (D3Q27) lattice for FPC-LBM for hydrodynamics and three-dimensional, fifteen velocity (D3Q15) lattice for FPC-LBM for temperature field. In each case, by matching the variations in the discrete central moments under collision to the corresponding continuous central moments of the FP collision operator, we will construct the equivalent central moment attractors. Since the number of collision invariants of the collision operators for fluid motions and energy transport in the DDF framework are different, it will be shown that the resulting attractors will be quite distinct. We will then validate these new LB schemes for buoyancy-driven thermal convective flows in square and cubic cavities over a wide range of Rayleigh numbers by comparing them against the respective benchmark numerical solutions. Finally, we will then demonstrate the advantages of using these FPC-LBMs for simulating mixed thermal convection at higher Peclet numbers than what can be achieved using the standard central moment LBMs using attractors based on the Maxwell distribution.

This paper is organized as follows. In the next section (§ 2), we will summarize the target macroscopic governing equations for thermal convective flows, which is followed by a highlight of the two continuous Boltzmann-type kinetic equations for representing hydrodynamics and energy transport in § 3. Then, § 4 discusses the 2D continuous formulations and the discrete LB schemes based on the Fokker-Planck central moments-based collision models, while their extensions to 3D are presented in § 5. Subsequently, the numerical case studies for the validation of the 2D and 3D FPC-LBMs are elaborated in § 6. Finally, a comparative study of numerical stability of the FPC-LBMs and the prior central moment LBMs that use the Maxwell distribution are discussed in § 7; this is followed by a summary of the new developments and findings resulting from this paper in § 8. Additional attendant details of this work, including the implementation details of the FP-LBMs in 2D and 3D as well as the guidance on the selection of the model parameters, are collected together in various appendices.

Macroscopic Governing Equations of Thermal Convective Flows

In this paper, our main goal is to develop and investigate novel LB schemes based on Fokker-Planck central moment-based collision modeling for effective simulations of thermal convective flows; based on stability considerations, the approach will involve the use of dual distribution functions – one for obtaining the solutions of the hydrodynamic fields and another for solving the temperature field. In this regard, first, we briefly recall the attendant macroscopic governing equations. The mass and momentum equations of the fluid motions read as

∂ρ∂t+∇⋅(ρu)=0,(1)\frac{\partial\rho}{\partial t}+\nabla\cdot(\rho\boldsymbol{u})=0, \tag*{(1)}

and

∂(ρu)∂t+∇⋅(ρuu)=−∇P+∇⋅σν+F,(2)\frac{\partial(\rho\boldsymbol{u})}{\partial t}+\nabla\cdot(\rho\boldsymbol{u}\boldsymbol{u})=-\nabla P+\nabla\cdot\boldsymbol{\sigma}_{\nu}+\boldsymbol{F}, \tag*{(2)}

where ρ\rho is the fluid density and u\boldsymbol{u} is the bulk fluid velocity. Moreover, PP is the fluid pressure, σν\boldsymbol{\sigma}_{\nu} is the viscous stress tensor, and F\boldsymbol{F} is an external body force (e.g., the buoyancy force). On the other hand, the temperature field satisfies the following energy transport equation that formally has the form of a convection-diffusion equation (CDE):

∂T∂t+∇⋅(uT)=∇⋅(α∇T),(3)\frac{\partial T}{\partial t}+\nabla\cdot(\boldsymbol{u}T)=\nabla\cdot(\alpha\nabla T), \tag*{(3)}

where TT is the local temperature, and α=k/(ρCp)\alpha=k/(\rho C_p) represents the thermal diffusivity, where kk is the thermal conductivity and CpC_p is the specific heat at constant pressure. Note that since any other scalar ϕ\phi (such as the concentration field) also obeys the same CDE as above for the temperature field, the novel collision models and LB schemes discussed in this work are directly applicable in such other cases.

Double Distribution Functions-Based Continuous Boltzmann-type Kinetic Equations using for Thermal Convective Flows

In general, the continuous Boltzmann equation, which evolves a single particle density distribution function, is a more fundamental representation of the thermo-hydrodynamic processes at a kinetic level than the macroscopic governing equations given in the previous section. It can be written as

∂f∂t+ξ⋅∇f=(δfδt)collBoltz+(δfδt)forcing,(4)\frac{\partial f}{\partial t}+\boldsymbol{\xi}\cdot\nabla f=\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}^{\mathrm{Boltz}}+\left(\frac{\delta f}{\delta t}\right)_{\mathrm{forcing}}, \tag*{(4)}

where f=f(x,ξ,t)f=f(\boldsymbol{x},\boldsymbol{\xi},t) is the density distribution function at time tt with the position and the particle velocity represented by x\boldsymbol{x} and ξ\boldsymbol{\xi}, respectively. Here, the left hand side of Eq. (4) accounts for changes in the distribution function ff due to particle advection, sometimes referred to as streaming. On the other hand, the particles interact through collision that redistributes their velocities, while conserving the mass, momentum and energy of the interacting particle pair, and they are taken into consideration in the time variation in ff through the Boltzmann’s binary collision term (δfδt)collBoltz\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}^{Boltz} given by (see e.g., [51])

(δfδt)collBoltz=∫wb(f′f1′−ff1) db dε dξ1.(5)\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}^{Boltz}=\int wb(f'f_1'-ff_1)\,db\,d\varepsilon\,d\xi_1. \tag*{(5)}

Here, the distribution functions of the particle pairs pre- and post-collision are represented by f≡f(x,ξ,t)f\equiv f(\boldsymbol{x},\boldsymbol{\xi},t), f1≡f(x,ξ1,t)f_1\equiv f(\boldsymbol{x},\boldsymbol{\xi}_1,t), and f′≡f(x,ξ′,t)f'\equiv f(\boldsymbol{x},\boldsymbol{\xi}',t), f1′≡f(x,ξ1′,t)f_1'\equiv f(\boldsymbol{x},\boldsymbol{\xi}_1',t), respectively, with w=∣ξ−ξ1∣w=|\boldsymbol{\xi}-\boldsymbol{\xi}_1| being the relative speed, ε\varepsilon the azimuthal collision angle, and bb the impact parameter. These are accounted for in the first term in the right hand side of Eq. (4). Moreover, if the fluid is subjected to an external body force F\boldsymbol{F}, the resulting acceleration alters the distribution function ff in a manner represented through the second term (δfδt)forcing\left(\frac{\delta f}{\delta t}\right)_{\mathrm{forcing}} on the right hand side of Eq. (4) given by

(δfδt)forcing=−Fρ⋅∇ξf,(6)\left(\frac{\delta f}{\delta t}\right)_{\mathrm{forcing}}=-\frac{\boldsymbol{F}}{\rho}\cdot\nabla_{\boldsymbol{\xi}}f, \tag*{(6)}

where ∇ξ\nabla_{\boldsymbol{\xi}} is the gradient in the velocity space. By taking various statistical moments of the distribution function with respect to the certain monomials of the particle velocity of zeroth, first and second order, the fluid mass density, momentum and temperature, respectively, can be recovered. However, on the one hand, the continuous Boltzmann equation is formidable to use since the particle velocity space is infinite and its collision term is a multidimensional integral; on the other hand, if the particle velocity space is discretized using a smaller number of velocities to construct a lattice Boltzmann method (LBM) while using a simpler tractable model for collision, the resulting computation of mass density, momentum and temperature altogether using a single distribution function is known to lead to numerical instabilities. Hence, practical considerations suggest the use of double distribution functions in LBM, where one of them is used to recover the hydrodynamic fields and the other the temperature (see Ref. [42]).

Thus, a modified continuous kinetic equation to compute the hydrodynamic fields, which recovers Eqs. (1) and (2), is given by

∂f∂t+ξ⋅∇f=(δfδt)coll+(δfδt)forcing,(7)\frac{\partial f}{\partial t}+\boldsymbol{\xi}\cdot\nabla f=\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}+\left(\frac{\delta f}{\delta t}\right)_{\mathrm{forcing}}, \tag*{(7)}

where the collision term satisfies the following constraints reflecting the mass and momentum collision invariants (unlike the Boltzmann’s collision term which also respects an additional invariant related to energy):

∫(δfδt)colldξ=0,∫(δfδt)collξdξ=0.(8)\int\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}d\xi=0,\qquad\int\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}\boldsymbol{\xi}d\xi=0. \tag*{(8)}

Then, the fluid density ρ\rho and bulk velocity u\boldsymbol{u} can be obtained by taking the zeroth and first moments, respectively, of the distribution function ff obtained from solving Eq. (7), which read as

ρ=∫f dξ,ρu=∫fξ dξ.(9)\rho=\int f\,d\xi,\qquad\rho\boldsymbol{u}=\int f\boldsymbol{\xi}\,d\xi. \tag*{(9)}

The second distribution function gg used to recover the energy transport equation (Eq. (3)) can be written as

∂g∂t+ξ⋅∇g=(δgδt)coll,(10)\frac{\partial g}{\partial t}+\boldsymbol{\xi}\cdot\nabla g=\left(\frac{\delta g}{\delta t}\right)_{\mathrm{coll}}, \tag*{(10)}

where the local temperature is obtained by taking the zeroth order moment of gg. That is,

T=∫g dξ.(11)T=\int g\,\mathrm{d}\boldsymbol{\xi}. \tag*{(11)}

Its collision term thus satisfies the following separate invariant:

∫(δgδt)colldξ=0.(12)\int\left(\frac{\delta g}{\delta t}\right)_{\mathrm{coll}}\mathrm{d}\boldsymbol{\xi}=0. \tag*{(12)}

Fokker-Planck (FP) Central Moments-Based Collision Modeling: 2D Formulations and LB Schemes

One of our objectives in this manuscript is to construct robust models for the essential features of the collision operators in Eqs. (7) and (10), viz., (δfδt)coll\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}} and (δgδt)coll\left(\frac{\delta g}{\delta t}\right)_{\mathrm{coll}} based on the Fokker-Planck central moment formulations and subsequently develop effective LB algorithms for thermal convective flows in a double distribution functions (DDF)-based framework, which will be discussed in what follows.

Construction of FP Central Moment Collision Operator: Fluid Motions

Since, as mentioned earlier, the collision term of the Boltzmann equation is computationally intractable, it is generally replaced with simpler models that retain the essential characteristics of the collision process. Many collision models have been constructed over the years; one popular model is the Bhatnagar-Gross-Krook (BGK) model [52] given by

(δfδt)collBGK=ωBGK(fM−f),(13)\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}^{\mathrm{BGK}}=\omega_{\mathrm{BGK}}(f^{\mathrm{M}}-f), \tag*{(13)}

where fMf^{\mathrm{M}} is the Maxwell distribution function written as

fM=ρ(2πcs2)N/2exp⁡[−(ξ−u)22cs2],(14)f^{\mathrm{M}}=\frac{\rho}{(2\pi c_s^2)^{N/2}}\exp\left[-\frac{(\boldsymbol{\xi}-\boldsymbol{u})^2}{2c_s^2}\right], \tag*{(14)}

and ωBGK\omega_{\mathrm{BGK}} is the characteristic frequency of collisions. In Eq. (14), which is essentially the Gaussian or normal distribution, ρ\rho and u\boldsymbol{u} are the fluid density and bulk velocity, respectively, and csc_s is the speed of sound, and NN is the number of spatial dimensions. As such, Eq. (13) models the effect of collisions as a relaxation process toward an equilibrium state defined in Eq. (14), and is a dramatically simplified approximation for the collision integral (Eq. (5)). When used in constructing LB methods via suitable discretization in a DDF-based framework, the BGK model leads to numerical instability issues especially in simulating flows with relatively low viscosities or thermal diffusivities (or, equivalently, high Reynolds numbers or Peclet numbers). Thus, we look for alternatives that can address these limitations.

Another simplified model for the collision term is the Fokker-Planck (FP) model [39]. While the FP equation was originally derived to model the Brownian dynamics involving drift and diffusion of particles, it is also considered as a model for the Boltzmann’s collision integral term when the particle pairs undergo grazing collisions with small velocity changes that commonly occur in scattering processes with inverse power interaction potentials and associated with large impact parameters (see e.g., Refs. [40, 53, 54]). The FP collision model can be represented as [39]

(δfδt)collFP=ωFP[∂∂ξi((ξi−ui)f)+Dij′∂2f∂ξi∂ξj].(15)\left(\frac{\delta f}{\delta t}\right)_{\mathrm{coll}}^{FP} = \omega_{FP}\left[\frac{\partial}{\partial\xi_i}\left((\xi_i-u_i)f\right)+D'_{ij}\frac{\partial^2 f}{\partial\xi_i \partial\xi_j}\right]. \tag*{(15)}

Here, Dij′D'_{ij} is the diffusion tensor parameter that is a function of second order central moments or the variance of the distribution function (which will be further elaborated in later sections), whereas the effective diffusion rate coefficient DijD_{ij} is dependent on the diffusion tensor parameter Dij′D'_{ij} via the FP relaxation frequency ωFP\omega_{FP} as Dij=Dij′ωFPD_{ij}=D'_{ij}\omega_{FP}. The FP collision model given in Eq. (15) represents the effect of collisions in terms of drift and diffusion of the particle distribution in the velocity space. The drift term (or the first term in the right hand of Eq. (15)) represents a process similar to that of dynamic friction and brings to light the idea that collisions act to eliminate gradients in the particle velocities of the fluid; on the other hand, the diffusion term (or the second term in the right hand of Eq. (15)) models the spread of the distribution function under collision via molecular chaos [53]. These mesoscopic processes of particle populations as modeled by the FP formulation, under coarse-graining, i.e., at larger length and time scales, correspond to the fluid dynamics of the bulk continuum as represented by the Navier-Stokes equations which can be established via the Chapman-Enskog multiscale expansions (see e.g., [33]). Interestingly, the FP collision model (Eq. (15)) maintains the quadratic non-linearity seen in Boltzmann’s collision integral (Eq. (5)) since uiu_i and Dij′D'_{ij} appearing in Eq. (15) could themselves be functions of the distributions, ff. We also note here that Eq. (15) is more general than that of the FP model introduced in [39] since we consider the diffusion parameter Dij′D'_{ij} to be a tensor rather than a scalar D′D'.

As discussed in our recent work [33], using Eq. (15) directly in constructing LB algorithms is cumbersome since it involves evaluating the derivatives of the distribution function in the velocity space, while they can be effectively and exactly represented by reformulating in terms of their central moments. To that end, we begin with a formal definition of the inner product of any two objects aa and bb as being the product of aa and bb, integrated over velocity space in 2D as

References

  1. [1]WJ Minkowycz, EM Sparrow, JY Murthy, and John P Abraham. Handbook of numerical heat transfer. 2009.
  2. [2]Manoj Kumar Tripathi and Kirti Chandra Sahu. Motion of an air bubble under the action of thermocapillary and buoyancy forces. Computers & Fluids, 177:58–68, 2018.arxiv.org/abs/1806.06491
  3. [3]Bashir Elbousefi, William Schupbach, Kannan N Premnath, and Samuel WJ Welch. Thermocapillary convection in superimposed layers of self-rewetting fluids: Analytical and lattice Boltzmann computational study. International Journal of Heat and Mass Transfer, 208:124049, 2023.arxiv.org/abs/2210.07501
  4. [4]Palas Kumar Farsoiya, Stéphane Popinet, Howard A Stone, and Luc Deike. Coupled volume of fluid and phase field method for direct numerical simulation of insoluble surfactant-laden interfacial flows and application to rising bubbles. Physical Review Fluids, 9(9):094004, 2024.DOI
  5. [5]Bashir Elbousefi, William Schupbach, Kannan N Premnath, and Samuel WJ Welch. Investigation of surfactant-laden bubble migration dynamics in self-rewetting fluids using lattice Boltzmann method. Physics of Fluids, 36(11), 2024.arxiv.org/abs/2408.00146
  6. [6]Lin Li, Pu Xu, Qihan Li, Zichao Yin, Runyuan Zheng, Jiafeng Wu, Jiajia Bao, Wenyu Bai, Huan Qi, and Dapeng Tan. Multi-field coupling particle flow dynamic behaviors of the microreactor and ultrasonic control method. Powder Technology, page 120731, 2025.DOI
  7. [7]Gaoan Zheng, Pu Xu, Tong Wang, and Qing Yan. Study on the bubble collapse characteristics and heat transfer mechanism of the microchannel reactor. Processes, 13(1):281, 2025.DOI
  8. [8]Eric D Siggia. High Rayleigh number convection. Annual review of fluid mechanics, 26(1):137–168, 1994.
  9. [9]Richard H Pletcher, John C Tannehill, and Dale Anderson. Computational fluid mechanics and heat transfer. CRC press, 2012.
  10. [10]Joel H Ferziger, Milovan Perić, and Robert L Street. Computational methods for fluid dynamics. springer, 2019.
  11. [11]Wim A Mulder. A new multigrid approach to convection problems. Journal of Computational Physics, 83(2):303–323, 1989.
  12. [12]Nader Ben Cheikh, Brahim Ben Beya, and Taieb Lili. Benchmark solution for time-dependent natural convection flows with an accelerated full-multigrid method. Numerical Heat Transfer, Part B: Fundamentals, 52(2):131–151, 2007.
  13. [13]Nelson O Moraga, Marcelo A Marambio, and Roberto C Cabrales. Geometric multigrid technique for solving heat convection-diffusion and phase change problems. International Communications in Heat and Mass Transfer, 88:108–119, 2017.DOI
  14. [14]Benzi, Roberto, Sauro Succi, and Massimo Vergassola. The lattice Boltzmann equation: theory and applications. Physics Reports, 222(3):145–197, 1992.DOI
  15. [15]Shiyi Chen and Gary D Doolen. Lattice Boltzmann method for fluid flows. Annual review of fluid mechanics, 30(1):329–364, 1998.
  16. [16]Pierre Lallemand, Li-Shi Luo, Manfred Krafczyk, and Wen-An Yong. The lattice Boltzmann method for nearly incompressible flows. Journal of Computational Physics, 431:109713, 2021.DOI
  17. [17]Haowen Xi, Gongwen Peng, and So-Hsiang Chou. Finite-volume lattice Boltzmann method. Physical Review E, 59(5):6202, 1999.pubmed.ncbi.nlm.nih.gov/11969609
  18. [18]Dhiraj V Patil and KN Lakshmisha. Finite volume tvd formulation of lattice Boltzmann simulation on unstructured mesh. Journal of Computational Physics, 228(14):5262–5279, 2009.DOI
  19. [19]Ahad Zarghami, Stefano Ubertini, and Sauro Succi. Finite volume formulation of thermal lattice Boltzmann method. International Journal of Numerical Methods for Heat & Fluid Flow, 24(2):270–289, 2014.DOI
  20. [20]Weidong Li and Li-Shi Luo. Finite volume lattice Boltzmann method for nearly incompressible flows on arbitrary unstructured meshes. Communications in Computational Physics, 20(2):301–324, 2016.DOI
  21. [21]Xiaoyi He and Li-Shi Luo. Theory of the lattice Boltzmann method: From the Boltzmann equation to the lattice Boltzmann equation. Physical review E, 56(6):6811, 1997.DOI
  22. [22]Cyrus K Aidun and Jonathan R Clausen. Lattice-Boltzmann method for complex flows. Annual review of fluid mechanics, 42(1):439–472, 2010.DOI
  23. [23]Leila Jahanshaloo, Emad Pouryazdanpanah, and Nor Azwadi Che Sidik. A review on the application of the lattice Boltzmann method for turbulent flow simulation. Numerical Heat Transfer, Part A: Applications, 64(11):938–953, 2013.DOI
  24. [24]Qing Li, Kai Hong Luo, QJ Kang, YL He, Q Chen, and Q Liu. Lattice boltzmann methods for multiphase flow and phase-change heat transfer. Progress in Energy and Combustion Science, 52:62–105, 2016.DOI
  25. [25]Keerti Vardhan Sharma, Robert Straka, and Frederico Wanderley Tavares. Current status of lattice Boltzmann methods applied to aerodynamic, aeroacoustic, and thermal flows. Progress in Aerospace Sciences, 115:100616, 2020.DOI
  26. [26]Yue-Hong Qian, Dominique d’Humìeres, and Pierre Lallemand. Lattice BGK models for Navier-Stokes equation. Europhysics letters, 17(6):479, 1992.DOI
  27. [27]Dominique d’Humìeres. Generalized lattice-Boltzmann equations. Rarefied gas dynamics, 1992.
  28. [28]Dominique d’Humières. Multiple–relaxation–time lattice Boltzmann models in three dimensions. Philosophical Transactions of the Royal Society of London. Series A: Mathematical, Physical and Engineering Sciences, 360(1792):437–451, 2002.pubmed.ncbi.nlm.nih.gov/16214687
  29. [29]Martin Geier, Andreas Greiner, and Jan G Korvink. Cascaded digital lattice Boltzmann automata for high reynolds number flow. Physical Review E—Statistical, Nonlinear, and Soft Matter Physics, 73(6):066705, 2006.pubmed.ncbi.nlm.nih.gov/16907021
  30. [30]Yang Ning, Kannan N Premnath, and Dhiraj V Patil. Numerical study of the properties of the central moment lattice Boltzmann method. International Journal for Numerical Methods in Fluids, 82(2):59–90, 2016.arxiv.org/abs/1202.6351
  31. [31]Martin Geier, Andreas Greiner, and Jan G Korvink. A factorized central moment lattice Boltzmann method. The European Physical Journal Special Topics, 171(1):55–61, 2009.DOI
  32. [32]Martin Geier, Martin Schönherr, Andrea Pasquali, and Manfred Krafczyk. The cumulant lattice Boltzmann equation in three dimensions: Theory and validation. Computers & Mathematics with Applications, 70(4):507–547, 2015.DOI
  33. [33]William Schupbach and Kannan Premnath. Fokker-planck central moment lattice Boltzmann method for effective simulations of fluid dynamics. Fluids, 9(11):255, 2024.arxiv.org/abs/2410.09739
  34. [34]Nicolaas Godfried Van Kampen. Stochastic processes in physics and chemistry, volume 1. Elsevier, 1992.
  35. [35]Nelson Wax. Selected papers on noise and stochastic processes. Courier Corporation, 2014.
  36. [36]Linda E Reichl. A modern course in statistical physics. John Wiley & Sons, 2016.
  37. [37]James H Luscombe. Statistical mechanics: from thermodynamics to the renormalization group. CRC Press, 2021.DOI
  38. [38]Andrei Medved, Riley Davis, and Paula A Vasquez. Understanding fluid dynamics from Langevin and Fokker–Planck equations. Fluids, 5(1):40, 2020.DOI
  39. [39]JL Lebowitz, HL Frisch, and E Helfand. Nonequilibrium distribution functions in a fluid. The Physics of Fluids, 3(3):325–338, 1960.DOI
  40. [40]Carlo Cercignani. The Boltzmann Equation and Its Applications. Springer-Verlag, New York, 1988.DOI
  41. [41]Timm Krüger, Halim Kusumaatmaja, Alexandr Kuzmin, Orest Shardt, Goncalo Silva, and Erlend Magnus Viggen. The lattice Boltzmann method. Springer International Publishing, 10:978–3, 2017.
  42. [42]Xiaoyi He, Shiyi Chen, and Gary D Doolen. A novel thermal model for the lattice Boltzmann method in incompressible limit. Journal of computational physics, 146(1):282–300, 1998.DOI
  43. [43]Zhaoli Guo, Baochang Shi, and Chuguang Zheng. A coupled lattice BGK model for the Boussinesq equations. International Journal for Numerical Methods in Fluids, 39(4):325–342, 2002.DOI
  44. [44]Jia Wang, Donghai Wang, Pierre Lallemand, and Li-Shi Luo. Lattice boltzmann simulations of thermal convective flows in two dimensions. Computers & Mathematics with Applications, 65(2):262–286, 2013.DOI
  45. [45]Dario Contrino, Pierre Lallemand, Pietro Asinari, and Li-Shi Luo. Lattice-Boltzmann simulations of the thermally driven 2d square cavity at high Rayleigh numbers. Journal of Computational Physics, 275:257–277, 2014.DOI
  46. [46]Keerti Vardhan Sharma, Robert Straka, and Federico Wanderley Tavares. Natural convection heat transfer modeling by the cascaded thermal lattice Boltzmann method. International Journal of Thermal Sciences, 134:552–564, 2018.DOI
  47. [47]Fatma M Elseid, Samuel WJ Welch, and Kannan N Premnath. A cascaded lattice Boltzmann model for thermal convective flows with local heat sources. International Journal of Heat and Fluid Flow, 70:279–298, 2018.DOI
  48. [48]Linlin Fei, Kai Hong Luo, Chuandong Lin, and Qing Li. Modeling incompressible thermal flows using a central-moments-based lattice Boltzmann method. International Journal of Heat and Mass Transfer, 120:624–634, 2018.DOI
  49. [49]Farzaneh Hajabdollahi and Kannan N Premnath. Central moments-based cascaded lattice Boltzmann method for thermal convective flows in three-dimensions. International Journal of Heat and Mass Transfer, 120:838–850, 2018.arxiv.org/abs/1710.04751
  50. [50]Farzaneh Hajabdollahi, Kannan N Premnath, and Samuel WJ Welch. Cascaded lattice Boltzmann method based on central moments for axisymmetric thermal flows including swirling effects. International Journal of Heat and Mass Transfer, 128:999–1016, 2019.arxiv.org/abs/1806.10241
  51. [51]Gilberto M Kremer. An introduction to the Boltzmann equation and transport processes in gases. Springer Science & Business Media, 2010.DOI
  52. [52]Prabhu Lal Bhatnagar, Eugene P Gross, and Max Krook. A model for collision processes in gases. I. Small amplitude processes in charged and neutral one-component systems. Physical review, 94(3):511, 1954.DOI
  53. [53]Tamas I Gombosi. Gaskinetic theory. Number 9. Cambridge University Press, New York, 1994.DOI
  54. [54]Richard L Liboff. Kinetic theory: classical, quantum, and relativistic descriptions. Springer Science & Business Media, 2003.
  55. [55]Lin-Yuan Chen, Nigel Goldenfeld, and Yoshitsugu Oono. Renormalization group and singular perturbations: Multiple scales, boundary layers, and reductive perturbation theory. Physical Review E, 54(1):376, 1996.pubmed.ncbi.nlm.nih.gov/9965081
  56. [56]Nigel Goldenfeld. Lectures on phase transitions and the renormalization group. CRC Press, 2018.DOI
  57. [57]Gabriel T Csanady. Turbulent diffusion in the environment, volume 3. Springer Science & Business Media, 2012.
  58. [58]Kannan N Premnath and Sanjoy Banerjee. Inertial frame independent forcing for discrete velocity Boltzmann equation: implications for filtered turbulence simulation. Communications in Computational Physics, 12(3):732–766, 2012.DOI
  59. [60]Eman Yahia and Kannan N Premnath. Central moment lattice Boltzmann method on a rectangular lattice. Physics of Fluids, 33(5), 2021.DOI
  60. [61]Graham de Vahl Davis. Natural convection of air in a square cavity: a bench mark numerical solution. International Journal for numerical methods in fluids, 3(3):249–264, 1983.
  61. [62]M Hortmann, M Perić, and G Scheuerer. Finite volume multigrid prediction of laminar natural convection: bench-mark solutions. International journal for numerical methods in fluids, 11(2):189–207, 1990.DOI
  62. [63]HN Dixit and V Babu. Simulation of high Rayleigh number natural convection in a square cavity using the lattice Boltzmann method. International journal of heat and mass transfer, 49(3-4):727–739, 2006.DOI
  63. [64]Toru Fusegi, Jae Min Hyun, and Kunio Kuwahara. A numerical study of 3D natural convection in a cube: effects of the horizontal thermal boundary conditions. Fluid dynamics research, 8(5-6):221, 1991.DOI
  64. [65]Peng Wang, Yonghao Zhang, and Zhaoli Guo. Numerical study of three-dimensional natural convection in a cubical cavity at high Rayleigh numbers. International Journal of Heat and Mass Transfer, 113:217–228, 2017.DOI
  65. [66]Y. S. Tian and T.G. Karayiannis. Low turbulence natural convection in an air filled square cavity: part i: the thermal and fluid flow fields. International Journal of Heat and Mass Transfer, 43(6):849–866, 2000.
  66. [67]S Mergui, F Penot, and JL Tuhault. Experimental natural convection in an air-filled square cavity at Ra= 1.7×10^9. RAWM Henkes & CJ Hoogendoorn (eds.), pages 97–108, 1993.
  67. [68]RJ Krane. Some detailed field measurements for a natural convection flow in a vertical square enclosure. In Proceedings of the First ASME-JSME Thermal Engineering Joint Conference, 1983, volume 1, pages 323–329, 1983.
  68. [69]SM Bilski, John R Lloyd, and KT Yang. An experimental investigation of the laminar natural convection velocity field in square and partitioned enclosures. In International Heat Transfer Conference Digital Library. Begel House Inc., 1986.DOI
  69. [70]Patrick Le Quéré and Masud Behnia. From onset of unsteadiness to chaos in a differentially heated square cavity. Journal of fluid mechanics, 359:81–107, 1998.DOI
  70. [71]DG Briggs and DN Jones. Two-dimensional periodic natural convection in a rectangular enclosure of aspect ratio one. ASME Journal of Heat Transfer, 107:850–4, 1985.DOI
  71. [72]E Tric, G Labrosse, and M Betrouni. A first incursion into the 3d structure of natural convection of air in a differentially heated cubic cavity, from accurate numerical solutions. International Journal of Heat and Mass Transfer, 43(21):4043–4056, 2000.DOI
  72. [73]Ao Xu, Le Shi, and Heng-Dong Xi. Lattice Boltzmann simulations of three-dimensional thermal convective flows at high Rayleigh number. International Journal of Heat and Mass Transfer, 140:359–370, 2019.DOI

Paper details

vlnc:W_uhgkate9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents